Range construction: getting help from a GTO solver

great - i tried to use PioSolver myself briefly after my snowie license ran out, definitely going to start mapping some ranges out as my study target this month OOP ranges / 3B ranges

Hey, definitely nice one!!

I have a couple of questions:
1) in gto approach should we care about Villain's range? or only how to play our range? Cause I thought it was the second.
For example, if you give the Villain a cold calling range of only AA is there any way to have balanced stats?

2) this program suggests to x/r a lot. Is that regardless the frequency Villain bets? Or assumes the frequency it recommends for the guy ip... and only then we should x/r that range...?

3) I don't understand also the strategy regarding the guy ip. Is he betting 1/3 cause the mp x/r a lot?

If this is an equilibrium suggested strategy for the 2 opponents - positions and the given ranges, then of course it makes sense, but on the other hand nobody makes money But if one deviates from this strategy and the other keep playing the same then the one who deviates looses? So if you keep x/r the range the program suggests and I bet only trips or better, you will still be making money ?

In this instance i believe we create a GTO plan based on our opponents range. Almost like a "GTO" pair, our strategy is paired to his strategy so the solver creates both strategies to be unexploitable to each other.

As we don't know his true range we could play this assumed strategy to protect ourselves from being exploited until we can adjust or factor population reads into our decisions. It also allows us to understand the principals of balancing to be able to apply them vs different opponants

As for the last point I think we make the money because when you check back so much you give us free cards but I guess it starts to make bigger mistakes for us raising you with trips or better not sure

Originally posted by asimos
Hey, definitely nice one!!

I have a couple of questions:
1) in gto approach should we care about Villain's range? or only how to play our range? Cause I thought it was the second.
For example, if you give the Villain a cold calling range of only AA is there any way to have balanced stats?

These programs only solve post flop situations, you have to tell them the pre flop ranges. Think about GTO this way: each player adjust and readjust to the other player infinite times, trying to exploit leaks until they no longer have a way of increasing their EV. GTO "cares" about opponent range. Like if I give in this example only AA to CO, what will happen is that MP is check/folding pretty much everything besides trips. And that's GTO given the ranges given. If pre flop GTO was possible here, CO would never have a range of only AA, that would never be the max EV play for pre flop. If you think about this as "trial and error" you get a more realistic sense of what's going on.

Originally posted by asimos
2) this program suggests to x/r a lot. Is that regardless the frequency Villain bets? Or assumes the frequency it recommends for the guy ip... and only then we should x/r that range...?

It does not assume anything. Let's take this example from the beggining, in a "trial and error" manner (which is what happens in these calculations), MP is cbetting most of his strong hands and some bluffs, CO adjusts and begins to exploit MP's weak check range with thin (big size) value bets and bluffs (betting a bunch versus missed on the flop). Now, MP has an incentive to start checking his strong hands since he gets more money in on the flop by check/raising than by cbetting those hands. Now CO has an incentive to start checking back a lot and fold less against MP cbets. And they go back and forth until they arrive to this equilibrium where no more adjustments are possible.

And this brings me to you last question:

Originally posted by asimos
If this is an equilibrium suggested strategy for the 2 opponents - positions and the given ranges, then of course it makes sense, but on the other hand nobody makes money But if one deviates from this strategy and the other keep playing the same then the one who deviates looses? So if you keep x/r the range the program suggests and I bet only trips or better, you will still be making money ?

If CO only bets vs missed with trips and checks the rest, he will give more EV to MP, even if MP does not change his strategy, for multiple reasons ( for example he lets MP realize too much equity by checking back too much), so although he reduces MP EV of check/raising he gives up EV on a lot of other tree branches. MP then could re-adjust and maximize even further his EV ( probablyby cbetting strong hands and not check/calling or check/raising almost anything). But this of course open him for exploitation.

Originally posted by asimos
3) I don't understand also the strategy regarding the guy ip. Is he betting 1/3 cause the mp x/r a lot?

I'm not sure. I think one of the reasons is that CO wants to protect against MP check/folding range (which have pretty decent equity everytime) but does not have much incetive to bet big because MP's top range is better than CO. If MP did not have so many combos of QQ+ CO would probably prefer bigger sizings.

People have a hard time to understand the concept of GTO, and I think it is a big mistake to call whatever this article describes GTO. It will add to the confusion, and many players will believe that GTO = Maximal exploitation, when, in fact, it is the opposite in a precise sense.

GTO does not care one wit about the opponents range. It is unexploitable, i.e. it works against every range, with an EV that is greater than or equal to zero (the latter only against another GTO player).

The content is good, but the terminology used is severely misleading.

To play exploitatively, we must be exploitable.

Originally posted by YohanN7
many players will believe that GTO = Maximal exploitation, when, in fact, it is the opposite in a precise sense.

GTO does not care one wit about the opponents range. It is unexploitable, i.e. it works against every range, with an EV that is greater than or equal to zero (the latter only against another GTO player).

When we notice that a player plays in a particular way, we can exploit that player by playing in a manner that counters it.
That is, we exploit his tendencies.

Lets look at a nit in the blinds. We steal wider. We may steal SO wide that the other blind starts to 3Bet us. He is thus exploiting OUR tendencies.

If I understand the GTO solver correctly, it looks only at postflop play, and attempts to create a balance between folds, bets, calls and raises, both in position and out of position.

Folding is an important part of playing optimally. Remember the old joke: "I want to move up stakes to where they respect my raises". They respect raises if you can fold.

Perhaps this little example does not represent "optimal" game theory play. But it is a terrific illustration of the process.

My worry is that if we play "perfectly" optimal play, we serve only to reduce our winrate. It seems to me that maximum winrate comes with being able to exploit everyone else's tendencies. So you bluff the nits, and fold to their raises. You never bluff calling stations, but bet, bet, bet your good hands. If you play at a mixed table, by changing your "style" depending on who you play against, everyone sees a well-balance player, when in fact you're just switching gears a lot.

Thanks, double2 -- nice video. I think that the video format works here, but in general the "classic" approach is better, if only because it dosn't take 20 minutes to read.

Cheers,
VS

I have no objections whatsoever - except for the terminology. What you describe is not GTO and should not be confused with it. It is the opposite of GTO, it is exploitation. GTO exploits every style, just not as much as it could do by abandoning GTO.

Cheers!

afaik
To get to GTO you still need to know the rules of the game i.e. you actually do know villain's range preflop i.e. theres 50(52) cards, 1225(1326) combinations ( or however much it is I guess you need to count card removal in true GTO solutions-that's why it's so complex) , and you get gto solution based on that.
These programs can't calculate even approximated gto from preflop as it's too complex though so they instead you "change" the rules of the game, simplify it and try to get to GTO approximate solutions based on the "changed rules" - so you input his range and your range and from that point on it estimates approximate gto solutions given these asymmetric ranges (and from that point on it does not care about opponent's range indeed)

I guess it is acceptable to put it that way. If the program can find a situation (two strategies since ranges and position differ) where no player can (unilatery) change his postflop strategy for the better, then it is an equilibrium, and it is GTO for that particular "set of rules (or combination of ranges)".

Originally posted by VorpalF2F
To play exploitatively, we must be exploitable.

Originally posted by YohanN7
many players will believe that GTO = Maximal exploitation, when, in fact, it is the opposite in a precise sense.

GTO does not care one wit about the opponents range. It is unexploitable, i.e. it works against every range, with an EV that is greater than or equal to zero (the latter only against another GTO player).

My worry is that if we play "perfectly" optimal play, we serve only to reduce our winrate. It seems to me that maximum winrate comes with being able to exploit everyone else's tendencies. So you bluff the nits, and fold to their raises. You never bluff calling stations, but bet, bet, bet your good hands. If you play at a mixed table, by changing your "style" depending on who you play against, everyone sees a well-balance player, when in fact you're just switching gears a lot.

Thanks, double2 -- nice video. I think that the video format works here, but in general the "classic" approach is better, if only because it dosn't take 20 minutes to read.

Cheers,
VS

Yeah I love seeing Pedro's beautiful mug, and this is awesome, and I'd love it even more if there'd be a paragraph on the theory + the hand
and then jump into fiddling with the program in the video as copy pasting images from it etc. would take too long, and it's great to see how it works in action. Best of both worlds

To the application at micros - yeah, you will cut your EV tons if you stop exploiting people. An example of the top of my head: Cbetting way too much to a massive size oop compared to what these programs suggest still makes a lot of money as people make blatant predictable mistakes.

But it's really fun getting these suggestions from the program, like I have a friend who runs these for me, I give him my range, sizing, ask him if my 2 cards are in my cbet range. And he runs it through, replies "no" "in fact you cbet 0.56% of the time with that range and sizing on that board" - you essentially get the optimal(ish, however these are accurate) solution, and then try to understand the relationship between balanced play and what you/people are actually doing so you know hot to exploit them.

Pretty much all regs cbet waytoo big and too much at 50 and lower for example, and before I've seen these suggestion I assumed that's normal as I do it too (for a good reason vs the population pool) and didn't try to exploit it - knowing the optimal approximation lets you exploit people better and spot their imbalanced is what I'm trying to say.

Or the turn 3bet pot thing - gto solver suggest whacky stuff like check shoving, overbet shovelling after betting 50% OTF quite a bit - even if you won't use those understanding why it suggests that and the relationship between standard line and those suggested again lets you adjust to people better

Originally posted by YohanN7
I guess it is acceptable to put it that way. If the program can find a situation (two strategies since ranges and position differ) where no player can (unilatery) change his postflop strategy for the better, then it is an equilibrium, and it is GTO for that particular "set of rules (or combination of ranges)".

yeah that's what they are attempting to do (approximating - doubt they find the solutions perfectly. And betsizes are fixed, simplifying the game eve further), pretty sure that's what Doube2 meant in terms of villain's ranges mattering and it's seen in the video as well when the range changes when it's BvB.

And if I get this right - actually curious about this
1) Since there is always range asymmetry, one player's EV is higher than the other ones just by getting with his range to a specific flop vs opponents range)
2) If what the program suggests is indeed GTO solution for both, both playing their ranges the way program suggests won't increase or decrease the EV advantage of one of the players.
3) If one of the players deviates from the strategy, he should lose money (not sure about this condition, but if there is one nash equilibrium, deviating from it should lower your EV as you are e.g. either bluffing too much or not enough; calling too much or folding too much etc?)

Well, if I pass an idea than GTO = max exploitation, it was not my intention. But I think it's useful to think about GTO and Nash Equilibrium as two players that adjust and readjust to each other until no one can increase EV by playing different, and if they do, they will lose EV.

Yes, these programs are very far from a real GTO solution, because real GTO would include a multitude of different betsizes and of course would have a pre flop component (with again a multitude of betsizes). What these programs do is that they give you a GTOish solution based on the "rules" you give them (like pre flop ranges and betsizes available).

Originally posted by YohanN7
I guess it is acceptable to put it that way. If the program can find a situation (two strategies since ranges and position differ) where no player can (unilatery) change his postflop strategy for the better, then it is an equilibrium, and it is GTO for that particular "set of rules (or combination of ranges)".

Yes, it's exactly that what it does. It ends up being a subset of GTO, we can play very far from GTO in a previous street and still play "GTOish" in the next.

But I'm not sure what do you suggest I do different or say different. Maybe I didn't express myself well enough.

Originally posted by LemOn36

Originally posted by YohanN7
I guess it is acceptable to put it that way. If the program can find a situation (two strategies since ranges and position differ) where no player can (unilatery) change his postflop strategy for the better, then it is an equilibrium, and it is GTO for that particular "set of rules (or combination of ranges)".

yeah that's what they are attempting to do (approximating - doubt they find the solutions perfectly. And betsizes are fixed, simplifying the game eve further), pretty sure that's what Doube2 meant in terms of villain's ranges mattering and it's seen in the video as well when the range changes when it's BvB.

And if I get this right - actually curious about this
1) Since there is always range asymmetry, one player's EV is higher than the other ones just by getting with his range to a specific flop vs opponents range)
2) If what the program suggests is indeed GTO solution for both, both playing their ranges the way program suggests won't increase or decrease the EV advantage of one of the players.
3) If one of the players deviates from the strategy, he should lose money (not sure about this condition, but if there is one nash equilibrium, deviating from it should lower your EV as you are e.g. either bluffing too much or not enough; calling too much or folding too much etc?)

The 3 are correct IMO.

3) If one of the players deviates from the strategy, he should lose money (not sure about this condition, but if there is one nash equilibrium, deviating from it should lower your EV as you are e.g. either bluffing too much or not enough; calling too much or folding too much etc?)

One question is if there is only one equilibrium and the answer to it is not obvious to me.

The other question is how the program converges to its solution. In numerical simulations you can end up to different results (if you don't prove that the solution is unique) for different initial conditions.

Because by doing try and error, for all possible turn and rivers and for all possible actions and for all hands in ranges shouldn't be a trivial task.

So not sure if we can trust a program if we do not know how exactly it works.

Maybe one should try different gto solvers for the same ranges. And then if they give the same result that could be an indicator that calculations are probably correct.

We know how they work. And it's no trivial task. The average calculation takes 6-8 GB of RAM and 3-10min of waiting, depending on the program and on the computer. And they give you a distance from de Nach Equilibrium, the more you wait, the more accurate it is. Usually <0.5% it's already good enough.

Nice video,

I was hoping to see the flop decision locked as hero played (perhaps with some similar hands) and then see how good the turn and river decisions were.

@asimos, please be a bit more careful when you quote something, and then edit it. What you attribute to me is not written by me.

Originally posted by jules97
Nice video,

I was hoping to see the flop decision locked as hero played (perhaps with some similar hands) and then see how good the turn and river decisions were.

I'll try to do that, I know that this is possible to do, but never done it before.

Originally posted by YohanN7
@asimos, please be a bit more careful when you quote something, and then edit it. What you attribute to me is not written by me.

Oops, really sorry, I edit it

Nice one, this one.